The TUI Dashboard: Watch Your AI's Brain Form Memories

AI systems are often black boxes. You put data in, results come out, and what happens in between is a mystery.

Shodh Memory includes something different: a Terminal User Interface (TUI) that lets you watch your AI's memory system in real-time. See memories form. Watch associations strengthen. Browse the knowledge graph. Explore indexed codebases.

Why a TUI?

Three reasons:

• Debugging: When the AI doesn't recall something you expect, you can see why. Is the memory stored? Is the association weak? Was it decayed?
• Understanding: Watching Hebbian strengthening in real-time builds intuition about how the system learns.
• Confidence: Seeing the memory system work builds trust in the system.

Plus, it looks cool.

The Main Dashboard

The dashboard is your command center for memory introspection:

The dashboard shows:

• Memory List: All stored memories, grouped by type (Learning, Decision, Error, Context)
• Recent Activity: Real-time log of memory operations
• Stats: Quick metrics—total memories, edges, LTP connections, storage size
• Session Context: What's currently active across different users/sessions

Real-Time Activity Feed

Watch operations as they happen:

14:23:01  REMEMBER   "JWT refresh token implementation requires..."
14:23:01  EMBED      384-dim vector generated (47ms)
14:23:01  NER        Entities: [JWT, refresh token, implementation]
14:23:01  EDGE       New edge: jwt ↔ authentication (strength: 0.15)
14:23:01  EDGE       Strengthened: auth ↔ security (+0.12 → 0.67)

Hebbian Learning Visualization

When memories are retrieved together, their connection strengthens:

RECALL query: "authentication security"
  Retrieved: [JWT tokens, Session handling, Password hashing]

  Edge updates:
    jwt ↔ session:    0.45 → 0.52 (+0.07)
    jwt ↔ password:   0.23 → 0.31 (+0.08)

  Co-activation count: jwt↔session = 8 (LTP threshold: 10)

After enough co-activations, connections become permanent (Long-Term Potentiation).

Knowledge Graph Explorer

Navigate the association graph visually:

The graph view shows:

• Nodes: Entities and memory clusters
• Edges: Connections with strength indicated by thickness
• LTP Edges: Permanent connections highlighted
• Activation: Recently accessed nodes glow

Projects and Todos View

The TUI also provides a complete view of your task management:

This view shows:

• Projects: Hierarchical project structure with sub-projects
• Todos: Tasks organized by status (backlog, todo, in_progress, done)
• Codebase Status: Which projects have indexed codebases
• File Browser: Navigate indexed files with related memories

Running the TUI

# Install the TUI binary
cargo install shodh-memory-tui

# Run it
shodh-tui

# Connect to a specific server
shodh-tui --server http://localhost:3030

# Watch a specific user's memories
shodh-tui --user claude-code

Keyboard Navigation

Debugging with the TUI

"Why didn't Claude recall this memory?"

1. Open the TUI, find the memory in the list
2. Check its importance score (too low?)
3. Check its last access time (decayed?)
4. Check connected edges (weak associations?)
5. Check entity overlap with query (no matching entities?)

"Why are these memories connected?"

1. Open the knowledge graph view
2. Find the edge between them
3. Check activation count (how many times co-retrieved?)
4. Check if LTP (permanent or temporary?)

The Value of Transparency

AI memory systems are only useful if you can trust them. Black-box storage that might or might not recall things correctly isn't helpful.

The TUI provides transparency: see exactly what's stored, watch learning happen in real-time, debug retrieval issues visually, understand the system's behavior.

This isn't just a debugging tool. It's a trust-building tool. When you can watch the AI's brain form connections, you understand why it behaves the way it does.